01
OpenAI sponsors Rosalind access for U.S. university labs
OpenAI announced the Rosalind Academic Plan, offering eligible U.S. university labs up to 30 Business Standard seats free for up to 12 months. Access to GPT-Rosalind, its specialized life-sciences model, requires trusted-access approval.
- Czech universities and corporate research teams should not budget on receiving this subsidy because the announced eligibility covers U.S. university labs.
- Life-sciences teams assessing GPT-Rosalind should evaluate the accompanying Rosalind workbench against specific medicinal chemistry, genomics or wet-lab workflows.
- Procurement teams should confirm trusted-access requirements and regional availability before scheduling a pilot.
Source — OpenAI Life sciencesResearch access 02
OpenAI agrees to restrict use of disputed Apple information
The Verge reported that OpenAI agreed not to use proprietary Apple information allegedly obtained by former Apple employees who joined OpenAI. OpenAI also agreed to provide forensic images of drives belonging to 14 employees while the litigation continues.
- Czech buyers should ask how OpenAI separates disputed third-party information from product development and customer environments.
- Legal teams should review vendor confidentiality commitments and intellectual-property protections rather than assume this agreement resolves the underlying allegations.
- Enterprises recruiting AI specialists from competitors should require onboarding checks that prevent staff from bringing confidential files into internal AI projects.
03
Reuters examines limits on independent AI safety reporting
Reuters reported that OpenAI and Anthropic sought significant control over what external evaluators could publish about their systems. OpenAI said it was not seeking approval rights over entire reports.
- Czech procurement teams should request evaluation contracts or summaries showing whether vendors can delay, redact or block publication.
- OpenAI said evaluators can disclose major redactions, so buyers should check those disclosures before relying on published safety findings.
- Teams deploying AI in regulated or high-impact workflows should require use-case-specific testing rather than treat a vendor’s external review as sufficient assurance.
Source — Reuters AI safetyProcurement 04
OpenAI reports employee arrest over alleged conference-room recording
The Verge reported that authorities arrested an OpenAI employee over allegations that he planted an encrypted recording device in an executive conference room. OpenAI said it had no indication that anyone accessed or exfiltrated model weights, source code or customer data.
- Czech enterprise security teams should include physical access and insider-risk controls in AI-vendor assessments, alongside cloud security.
- Buyers should confirm contractual incident-notification triggers and evidence-sharing obligations rather than treat the reported absence of data exposure as a completed forensic finding.
- OpenAI said it used its models to help identify the suspect, but buyers considering similar investigations should require human review and conventional forensic evidence.